Originally posted by: Ziggym

Hello everyone,

Got some theoretical ponderings I wanted to share and get some input on.

Lets say we to process 2 million messages (small ones, quick transactions) and have a rule like : 

1. IF(PRESENT(A) & ABSENT(B), "X")

Would it make any impact on performace to write this as :

2. IF(PRESENT(A),IF(ABSENT(B),"X"))

I realise I adding an IF to the rule but also cutting one part of the condition in case PRESENT(A) is false.

So to clear it up will option 2 make the processing of the rule slower (if both conditions needs to be tested) compared to rule 1 (ie do the & add a IF processing as well so its basically the same thing)

(I do realize that if condition A is not met rule 2 skips condition B and in those occurences should be quicker. A second path would ofc be to analyse the bulk of the messages and adjust the condition to whatever is most common.)

What's the verdict ? (if it all makes any sense)

Regards

Roberth

Originally posted by: Arun Ramamurthy

I wouldn't expect much of a difference between those two . From what I understand , there's some baggage that goes with an "IF" (we expect a condition, a "then" clause, and an "else" clause), but I would doubt that any of that makes a noticeable difference in performance in this situation. There may be some small differences as you indicate depending of the success of the condition , but that may not be noticeable is what I would think.

Originally posted by: Ziggym

Hi Arun,

Your right. There is a small difference that works to the case 2 advantage when the first condition isnt meet.

with a test set that causes 50% hits to the condition's case 1 is actually slightly quicker but to small to be taken any notice to.

Done a few unscientific test with input of 2 million records and 50 runs each and average difference to case 1's advantage is .03 seconds and well within faulty margins. The more misses on PRESENT(A) the quicker processing of case 2 is but its also marginal. 2 million inputs with no hits makes a difference of about .5 seconds in processing so not much to fuzz about.

There are other ways to improve performance that gives more back in regards to performance.

// Roberth

Originally posted by: GKearney

Hi,

The following Technote contains information about 'How to save memory and improve performance in WebSphere Transformation Extender Maps and systems'.

http://www-01.ibm.com/support/docview.wss?uid=swg21245727

I hope this information is of help to you.